Should-Cost Modelling for Machined Parts: The ₹855 Machine Hour, the 2.75:1 Buy-to-Fly That Turns ₹1,419 of Plate into a ₹1,266 Bracket, and the 61-Piece Point Where Casting Beats Billet

A quote is not a price list lookup, it is a cost model with a margin bolted on top, and almost every losing job in a job shop is lost at the modelling stage…

·

Should-Cost Modelling for Machined Parts

Ask ten job shops to quote the same bracket and you will get numbers spanning 4:1. Ask each of them what their machine hour costs and most will name a figure they inherited from a rate card written when their VMC was new, their power tariff was ₹6.20, and their operator earned ₹14,000 a month. The spread is not greed and it is not efficiency — it is that most quotes are built from a remembered rate rather than a cost model, and a remembered rate drifts in one direction only.

Five numbers frame this subject.

The first is ₹855. That is what an hour of a ₹45 lakh three-axis VMC actually costs a two-shift Indian shop in 2026, once depreciation, cost of capital, floor space, power, maintenance, consumables, a loaded operator, tooling, 25% overhead and a 15% contingency uplift are all counted. The market conversation in Coimbatore puts a VMC hour at ₹550 and reports buyers paying ₹250. A shop quoting ₹250 is not competing; it is liquidating a machine over five years and calling the proceeds profit.

The second is 2.75. A 190 × 130 × 38 mm plate weighs 2.534 kg. The finished bracket weighs 0.92 kg. The buy-to-fly ratio — the single number that separates a material line from a material problem — is 2.75:1, so you buy ₹1,419 of 6061-T6 and recover ₹153 as chips. Material is 52.3% of the finished price, which is why a 10% ingot move must produce a 5.23% price move and why shops that hold prices for a year on aluminium work quietly go under.

The third is 61. Below 61 pieces, machining from billet is cheaper than paying for a casting pattern. Above it, the ₹65,000 of tooling has amortised itself and the casting route wins by an increasing margin. Almost no shop computes this crossover, which is why they quote billet at 500-off and lose the job to a foundry, and quote casting at 20-off and lose money.

The fourth is 4.2%. Setup is ₹6,100 on this part. At one piece it is 31% of the price; at ten, 18.1%; at sixty, 4.2%; at six hundred, 0.4%. The same drawing therefore has a legitimate price of ₹6,311 or ₹2,264 depending only on quantity. Quoting the wrong quantity band is the most common single error in job-shop estimating.

The fifth is 1.60%. That is the working-capital cost of 45-day payment terms at a 13% cost of capital — and 45 days is the statutory maximum under Section 15 of the MSMED Act, not a target. At 90 days it is 3.21%. No line on the quote captures it, so it comes straight out of the margin.

This guide runs in eleven sections:

1. The Stack: One Identity, Two Ways to Get It Wrong

A should-cost model is not a discounting exercise. It is a reconstruction of the supplier's cost from physics and prices, and it has a fixed order of operations:

where m_s is stock mass, c_m the material price per kg, \kappa the chip-recovery fraction, r_{mh} the machine hour rate, t_c the booked cycle time in hours, S the total setup cost, N the batch quantity, Y the first-pass yield, o the SG&A absorption rate and p the target profit margin.

The identity looks harmless. The trap is in the last line. Overhead is applied as a markup on cost — 25% of cost is the conventional absorption — but margin is a share of price, so it must be divided out, not multiplied in:

Markup applied to cost · Realised margin · Multiplier needed to achieve that margin

10% · 9.09% · 1.111

12% · 10.71% · 1.136

20% · 16.67% · 1.250

A shop that "adds 20%" believes it is earning 20%. It is earning 16.7%, and the 3.3 points it never collected are larger than the net margin of most job shops. This one-line error, repeated across a year of turnover, is the difference between a viable shop and a shop that is always busy and never profitable.

flowchart LR
    A[Drawing + tolerance] --> B[Process route]
    B --> C[Cycle time from MRR]
    B --> D[Setup hours]
    C --> E[Machine hour rate]
    D --> F[Setup / N]
    A --> G[Stock mass x BTF]
    G --> H[Material net of chip credit]
    C --> I[Tool life: Taylor]
    E --> J[Processing cost]
    F --> K[Direct cost]
    H --> K
    I --> K
    J --> K
    K --> L[Divide by yield]
    L --> M[x 1 + overhead]
    M --> N[Divide by 1 - margin]
    N --> O[Add freight, terms]
    O --> P[Quote]
    classDef proc fill:#eef2ff,stroke:#1e3a5f,stroke-width:2px,color:#111827;
    classDef money fill:#ecfdf5,stroke:#065f46,stroke-width:2px,color:#111827;
    class B,C,D,E,F,G,H,I proc;
    class K,L,M,N,O,P money;

2. The Machine Hour: Six Components and One Denominator

Every machine hour rate argument is really an argument about the denominator. The numerator is arithmetic; the denominator is a policy decision about how many hours you are willing to sell in a year, and it moves the rate more than any other input.

Take a ₹45,00,000 landed three-axis VMC — a 1,000 mm travel machine with a 12,000 rpm spindle, a 24-tool carousel and a probe, installed, commissioned and tooled. Depreciate over ten years to a 10% salvage value, and sell 3,600 spindle-on hours a year: two shifts, 250 working days, 16 hours nominal, derated to 72% for setup, maintenance, breakdowns and the jobs that never arrive.

Component · Formula · ₹/hour

Depreciation · (K-S)/(L\,H) = (45,00,000 − 4,50,000)/(10 × 3,600) · 112.50

Cost of capital · \bar{K}i/H = 24,75,000 × 0.11/3,600 · 75.60

Floor space · A f \times 12/H = 12 m² × ₹350/m²/month × 12/3,600 · 14.00

Power · P_e\lambda\tau = 18 kW × 0.45 × ₹8.50/kWh · 68.80

Maintenance · 0.04K/H = 4% of capital per year · 50.00

Consumables · coolant, way lube, filters, wipers · 25.00

Machine only · · 346.00

Operator (loaded) · ₹28,000 × 1.35 ÷ 200 productive hours · 189.00

Tooling allowance · carbide, at ₹60/hr · 60.00

Subtotal · · 595.00

Overhead · 25% — supervision, QA, stores, admin, rent · 148.75

Cost per hour · · 743.75

Contingency uplift, 15% — breakdowns, rework, hours that never sell · 743.75 × 1.15 · 855

Three things in that table contradict the usual shop-floor intuitions.

Electricity is not the cost driver. Power is ₹68.80 an hour, 8.0% of the rate. A tariff move from ₹7.00 to ₹11.00 per kWh — a 57% increase, the kind that generates a week of complaint — moves the rate from ₹838 to ₹884, which is 5.4%. Capital recovery and labour are the rate; energy is a rounding error with a public relations problem.

Capital is not the main driver either. The same model at ₹80 lakh of capital gives ₹1,122/hr and at ₹1.20 crore — a five-axis with a trunnion and a 40-tool magazine — gives ₹1,426/hr. A machine costing 2.7× more costs only 67% more per hour, because the expensive part of an hour is the hour, not the iron. That is also why a five-axis shop can undercut a three-axis shop on a complex part and still make money: it is selling fewer hours at a higher rate against a part that used to need four setups.

Utilisation is the driver. Hold the capital at ₹45 lakh and change only the denominator:

Billable hours/year · Machine-only ₹/hr · Fully loaded rate ₹/hr

2,400 (single shift) · 472 · 1,036

3,600 (two shifts) · 346 · 855

4,800 (three shifts) · 283 · 765

6,000 (near-continuous) · 245 · 710

A single-shift shop's hour is 21% more expensive than a two-shift shop's hour for identical iron, identical power and identical wages. The idle capacity is not free; it is prepaid, and it has to be recovered from the hours that do sell. This is the arithmetic behind the Coimbatore complaint that VMC work "should" fetch ₹550 and is being bought at ₹250: at ₹250 an hour against a ₹1,036 cost, the shop is not covering depreciation, let alone capital.

The operator ratio is the last lever. At one operator per machine the rate is ₹855; at two machines per operator it is ₹719. Sixteen percent of the rate sits in a staffing decision that most shops make on the basis of how annoying the part is to load.

One caution on comparing rates: a "machine rate" that excludes labour, a "shop rate" that includes it, and a "charge-out rate" that includes margin are three different numbers, and quoting one against the other produces false savings. Always ask which. The DFMA convention is a fully burdened charge-out rate, and that is the only one that can be compared with a supplier's price.

3. Cycle Time: The Only Line That Scales With Nothing

Material cost scales with the part. Setup scales inversely with quantity. Cycle time scales with nothing — it is the one line item that stays constant per piece whether you make ten or ten thousand, which is exactly why it deserves to be estimated from physics rather than from the last similar job.

The removal rate is the product of three numbers:

For a 16 mm three-flute carbide end mill in 6061-T6 at v_c = 400 m/min, f_z = 0.10 mm, z = 3, a_p = 6 mm and a_e = 8 mm: n = 7{,}958 rpm, v_f = 2{,}387 mm/min, and \dot{V} = 114 cm³/min. Derate heavily — a real part has corners, floors, thin walls, tool changes and a fixture that blocks half the approach — and 45 cm³/min is a defensible booking rate for roughing.

For the bracket, stock volume is 938.6 cm³ and the finished part displaces 340.7 cm³, so 597.9 cm³ must be removed:

Add 6 min of finishing (the surfaces that carry tolerance, not the ones that carry paint), 3 min of drilling and tapping for eight M6 and four Ø8 holes, and the cutting time is 22.3 min. Apply a 1.12 booking factor for tool changes, rapids, probing and the operator's walk to the machine: 25.0 min booked, or ₹356 of machine time at ₹855/hr.

3.1 Taylor's equation as a cost constraint

Taylor's relation is usually taught as tool-life trivia. In a quote it is a constraint that decides whether the shop makes or loses money on the job:

The exponent n is the whole story. For high-speed steel n \approx 0.125; for cemented carbide 0.25–0.33; for ceramic and CBN 0.4–0.5. Feed the carbide case into the ratio and the result is uncomfortable:

Speed change · n = 0.125 (HSS) · n = 0.25 (carbide) · n = 0.33 (ceramic)

+25% · life × 0.168 · life × 0.410 · life × 0.509

+100% · life × 0.004 · life × 0.063 · life × 0.122

Run a carbide tool 25% faster and you keep 41% of its life. Double the speed and you keep one-sixteenth. A programmer who shaves two minutes off a 25-minute cycle by pushing the speed has traded 2 minutes of machine time (₹28) for 59% of an edge. If that edge costs ₹950 and was cutting 30 parts, the trade costs ₹560 per part to save ₹28. This is not a hypothetical: it is the single most common way a shop convinces itself it is efficient while its tooling account grows faster than its revenue.

Tooling cost per part follows directly:

At ₹950 per edge and a 25-minute cycle: a 60-minute edge life gives ₹396/part, a 30-minute life ₹792/part, a 15-minute life ₹1,583/part. In 6061 aluminium, edge life is measured in hours and the tooling line is genuinely small — the ₹60/hr allowance in the rate card is adequate. In Ti-6Al-4V at 50–60 m/min, 20–30 minutes of edge life is normal, and the same part carries ₹800–1,600 of tooling that no ₹60/hr allowance will ever capture. Titanium quotes that go wrong almost always go wrong in the tooling line, not the machine line.

4. Setup and Lot Size: The Only Number That Is Not Linear

Setup is a fixed cost per lot: programming, fixture design and build, tool assembly, first-article measurement, and the non-recurring thinking that happens before the spindle turns. For the bracket — two setups, a soft-jaw fixture with a datum stop, 2.5 hours of CAM and 4.5 hours on the machine — the total is:

Amortised over a lot, the per-part contribution is S/N, and its derivative is -S/N^2: the savings from doubling quantity are large at small N and vanish at large N. That produces the price ladder that defines job-shop economics:

Quantity · Setup per part · Direct cost · Price per part · Setup share of price

1 · ₹1,958* · ₹4,810 · ₹6,311 · 31.0%

10 · ₹610 · ₹2,561 · ₹3,361 · 18.1%

60 · ₹102 · ₹1,844 · ₹2,419 · 4.2%

600 · ₹10 · ₹1,726 · ₹2,264 · 0.4%

\At one piece the fixture is not built and the first-article CMM report is not produced; setup falls to 1.5 h machine plus 0.75 h programming. The anodising lot minimum still applies.*

The same drawing, the same material, the same machine, the same operator: ₹6,311 or ₹2,264, a factor of 2.8, entirely from quantity. Any conversation about whether a quote is "expensive" that does not fix the quantity first is not a conversation about price.

Two consequences follow.

Never quote a small lot at a large-lot rate. A shop that prices a 10-off at the 600-off number is buying the work with its own cash. The ₹2,264 rate assumes 600 pieces of setup amortisation and, usually, a learning curve the shop has not yet climbed. Wright's law — C_n = C_1 n^{b} with b = \log(0.85)/\log 2 = -0.234 for an 85% curve — describes a cost decline that is earned over production, not available on day one. Quoting as if it were already earned is how shops end up with a portfolio of work that is profitable on paper and cash-negative in reality.

Quantity changes the process, not just the price. The make-vs-buy crossover for this bracket, comparing billet machining against an A356 sand casting with ₹65,000 of pattern and core-box tooling:

Quantity · Billet route (marginal) · Casting route (incl. tooling)

30 · ₹1,622 · ₹2,730

60 · ₹1,622 · ₹1,646

100 · ₹1,622 · ₹1,213

200 · ₹1,622 · ₹888

Sixty-one pieces. That is the entire decision. A shop that quotes billet at 500-off is more than twice the casting route and will lose to any foundry that can do the arithmetic; a shop that quotes casting at 20-off is 2.3× the billet price and will lose to every machine shop. The crossover is not a judgement call — it is a division.

5. Material and Buy-to-Fly: You Buy the Chips

Material is the line that is most often estimated from the wrong mass. The instinct is to price the finished part. The cost is in the stock:

For the bracket: stock 190 × 130 × 38 mm = 938.6 cm³ at 2.70 g/cm³ = 2.534 kg; finished mass 0.92 kg; BTF = 2.75. At ₹560/kg the gross material is ₹1,419, the 1.614 kg of chips recover ₹153 at ₹95/kg, and the net is ₹1,266 — 52.3% of the ₹2,419 price.

Part class · Typical buy-to-fly · What it means for the quote

Near-net casting or forging · 1.2–2.0 · You pay the process instead of the metal

Sawn blank, face and edge only · 1.5–2.5 · Allowance and nominal thickness drive everything

Simple pocketed plate part · 3–5 · The bread-and-butter job-shop part

Deep-pocketed housing or manifold · 5–8 · Cycle time usually costs more than the metal

Aerospace bracket from plate · 8–15 · Worth a hard look at a thinner nominal

Aerospace structural rib · 15–25 · Where a 6 lb part comes out of a 120 lb billet

Four points of practice:

Chip credit is real but small, and it is a discipline not a rate. Clean, segregated 6061 turnings fetch ₹80–110/kg in India; mixed or wet chips fetch half that or nothing. On this part the credit is 10.8% of the material bill, which is worth having and is not worth a business model. On titanium at BTF 12 the arithmetic inverts: 11 kg of chips per kg of part, and if those chips are mixed with steel turnings the revert value collapses and the credit disappears.

Thickness is a cost decision made on the drawing. A 38 mm plate where 34 mm would do is 4 mm × 190 × 130 × 2.70 g/cm³ = 267 g of aluminium that becomes chips at ₹560/kg — ₹150 per part, ₹9,000 across a 60-off, for a change that costs a CAD edit and a tolerance review.

Material price is the highest-beta input in the quote. At a 52.3% material share, a 10% ingot move requires a 5.23% price move just to stand still. Aluminium, copper and nickel do not respect annual price agreements, so either the quote carries a material escalation clause or the margin absorbs the move. Silence on the subject is a decision to absorb it.

The titanium inversion. Ti-6Al-4V bar at 10–20× the price of aluminium, at 50–150 SFM instead of 1,000–5,000, with 5–10× shorter tool life and 3–5× longer cycles, turns material from a pass-through into the dominant line. A part that is ₹1,266 of aluminium becomes ₹4,000–8,000 of titanium before a tool touches it. Shops that quote titanium by scaling the aluminium number by density produce quotes that are wrong by a factor of three in both directions.

6. Yield, Cpk and Scrap: The Multiplier Applies to Everything

Yield is applied to the whole accumulated cost, not to the material. That distinction matters more than the arithmetic suggests:

First-pass yield · Cost multiplier · What it looks like in a 60-off

90% · 1.111 · 6 parts scrapped, 11.1% added to every price

95% · 1.053 · 3 parts scrapped

97% · 1.031 · 1.8 parts scrapped

99% · 1.010 · 0.6 parts scrapped

A part scrapped at final inspection loses ₹1,844 — material, machining, setup amortisation, deburring, anodising and inspection all spent — not the ₹1,266 of aluminium that a shop is tempted to count. Scrapping late is expensive precisely because the value added is not recoverable.

Yield is not a number to be asserted; it is set by the process capability against the tolerance. For a two-sided tolerance with mean \mu and standard deviation \sigma:

which, evaluated on a centred process, gives:

Cpk · Scrap (ppm) · Scrap rate · Price multiplier

0.67 · 44,431 · 4.44% · 1.047

0.80 · 16,395 · 1.64% · 1.017

1.00 · 2,700 · 0.27% · 1.003

1.10 · 967 · 0.097% · 1.001

1.33 · 66 · 0.0066% · 1.000

1.50 · 7 · 0.0007% · 1.000

The convention quotes 63 ppm at Cpk 1.33 because it truncates at exactly 4σ; the exact normal integral gives 66 ppm. Either way the message is the same: capability above 1.33 buys nothing in scrap terms, and capability below 1.0 costs real money. The reason to specify a tight tolerance is function, not cost — and the reason a tight tolerance costs more is not the scrap rate, it is the process required to reach the capability: a finer finishing pass, a temperature-controlled room, a CMM check, a stress relief between roughing and finishing.

That is the tolerance staircase, and it is why two parts identical except for one ±0.05 mm bore can differ by 40% in price. The bore did not become harder to cut. It became harder to prove, and proving is a cost.

One caution on the Cpk table: it assumes the process is centred so that C_{pk} = C_p. A process with C_p = 2.0 running 1σ off centre has C_{pk} = 1.0 and 2,700 ppm — the same scrap as a far less capable but centred process. Mean shift, not spread, is the usual reason a shop's yield is worse than its capability study predicted.

7. Outside Processing: Where Small Lots Die

Anodising, plating, heat treatment, passivation, laser marking and stress relief are pass-through costs with a minimum charge that dominates small lots. The pricing basis matters, because the two common Indian bases differ by an order of magnitude at low volume:

Basis · Rate · Bracket at 135 in² · At 60-off

Per kg (tonnage contract, 15–25 µm Type II) · ₹150–250/kg · 0.92 kg → ₹138–230/part · ₹8,280–13,800/lot

Per area (job work) · ₹0.25–1.00/in² · 135 in² → ₹34–135/part · ₹2,025–8,100/lot

Minimum lot charge · ₹1,200–2,500/lot · — · binds below ~20 parts

The per-kg rate quoted by large anodisers assumes tonne-scale batches and a 50,000 kg minimum order; applying it to a 60-piece job is how a quote ends up 40% high on finishing. The per-area job-work rate is the correct basis at this quantity, and the lot minimum is what actually sets the price for anything under about twenty pieces.

Three rules for the outside-processing line:

  1. Do not apply the shop's margin to a pass-through. Overhead absorption on purchased services is legitimate; a 15% profit on an anodiser's invoice is not, because the shop added no value and the customer can see the market rate. Charge a handling percentage (5–10%) or charge nothing and recover it in processing time.
  2. Model the minimum charge explicitly. ₹2,500 of anodising on a 5-piece lot is ₹500 per part, which can exceed the machining cost. The quote must show it or the shop absorbs it silently.
  3. Lead time is part of the price. A ₹1,200 anodising bill with a 10-day turnaround on a job the customer needed in 5 days is not a saving. Outside processing is where delivery promises break, because the shop controls its own machines and does not control the anodiser's queue.

8. The Costs Nobody Quotes

Three costs sit outside the stack and are large enough to erase a normal margin.

8.1 Receivables

Payment terms are a loan the supplier makes to the buyer at its own cost of capital:

Terms · Cost at 13% WACC · Cost at 16% (working-capital loan)

30 days · 1.07% · 1.32%

45 days · 1.60% · 1.97%

60 days · 2.14% · 2.63%

90 days · 3.21% · 3.95%

120 days · 4.27% · 5.26%

A 12% margin quote at 90-day terms is a 8.8% margin quote. At 120 days it is 7.7%. The terms are negotiated after the price, so they are never priced — which is precisely why they are the most reliable margin leak in the business.

In India there is a statutory floor as well as an economic one. Section 15 of the MSMED Act, 2006 caps the agreed payment period at 45 days from acceptance; Section 16 makes a buyer who misses it liable for compound interest with monthly rests at three times the RBI bank rate; Section 23 disallows that interest as a deduction under the Income Tax Act. The remedy is real — an MSE with Udyam registration can file on the MSME Samadhaan portal and the Facilitation Council can award the principal plus statutory interest, with a 75% deposit required to appeal. Very few job shops use it, which is why 90- and 120-day terms persist in the supply chain even where they are unlawful.

8.2 Estimating overhead

Quoting is work, and the work is spent whether or not the job is won:

At 1.5 hours per quote, ₹900/hr of estimating time and a 20% win rate, every won job carries ₹6,750 of estimating cost before the first cut. At a 15% win rate it is ₹9,000. On the 60-off bracket — a ₹1,45,140 job — that is 4.7% of revenue, larger than the profit. A shop that wins 40% of its quotes carries ₹3,375 per job, half as much, for the same estimating effort. Win rate is a cost lever, not just a sales metric, and the fastest way to raise it is to stop quoting work the shop is not set up to do.

8.3 Rework and the second-op trap

Rework is not covered by the yield multiplier if it is discovered after despatch. A 3% return rate on a 12% margin product, where each return costs the full part value plus freight and a credit note, is a negative margin on the affected lots. The reserve belongs in the model as a percentage of revenue (1–3% is typical for a shop with stable processes; 5%+ signals a process problem that no price can fix).

9. The Worked Example

The bracket: 6061-T6, 190 × 130 × 38 mm stock, 0.92 kg finished, two setups, eight M6 and four Ø8 holes, 135 in² of Type II clear anodise, ±0.05 mm on two bores, 60 pieces.

Line · Basis · ₹/part

Material, gross · 2.534 kg × ₹560/kg · 1,419

Chip credit · 1.614 kg × ₹95/kg · −153

Material, net · · 1,266

Machining · 25.0 min × ₹855/hr · 356

Setup amortised · ₹6,100 ÷ 60 · 102

Deburr and wash · 4 min × ₹250/hr · 17

Anodise Type II clear · 135 in² × ₹0.45/in² · 61

Inspection · FAI ₹1,800 ÷ 60 + 800/hr in-process · 43

Direct cost · · 1,844

Yield · ÷ 0.97 · 1,901

SG&A · × 1.12 · 2,129

Cost · · 2,129

Price · ÷ 0.88 (12% margin) · 2,419

Sanity checks on that number. Material is 52.3% of price, machining 14.7%, setup 4.2%, outside processing 2.5%. A buyer benchmarking against ₹/kg would compute ₹2,419 ÷ 0.92 = ₹2,630/kg and conclude the part is expensive; a buyer benchmarking against machining hours would compute ₹2,419 ÷ 0.417 h = ₹5,800 per machine hour and conclude the same thing, wrongly. Neither metric survives contact with a should-cost model, which is the argument for building one.

And the failure case. Suppose the shop quotes ₹2,419 and then discovers the material was ₹615/kg rather than ₹560 and one part in twenty needs rework — a 7.4% cost overrun that no one would call a disaster:

₹10,740 on a ₹1,45,140 job is 7.4% of revenue, which is more than half the profit. Two ordinary errors, no catastrophe, and the job is worth less than the estimating time it consumed.

10. Reading a Quote

For buyers — the six questions that separate a real quote from a number:

  1. What quantity is this priced at, and what does the next quantity break look like? A quote without a quantity ladder hides whether you are paying for setup or for metal.
  2. Is the machine rate fully burdened, and what utilisation is it based on? A ₹500/hr rate implies either a fully depreciated machine, a single-shift denominator the shop has not disclosed, or a rate that excludes labour.
  3. What buy-to-fly did you assume? If the answer is "finished weight", the material line is understated by 2–8× and something else in the quote is carrying it.
  4. What yield is in the price, and what is the scrap rate on this feature historically? A quote assuming 99% on a ±0.02 mm bore is a quote that will come back as a variation order.
  5. Which operations are outsourced, at what lead time, and what is the lot minimum? The finishing line is where delivery dates are actually decided.
  6. What are the payment terms and is there a material escalation clause? These move the effective price by 2–4% and are almost never discussed at the quotation stage.

For shops, the same list in reverse: a quote that states quantity basis, material basis, lead time, exclusions and validity date is a quote that does not have to be renegotiated. On FabFlow, a quotation carries price, estimated completion date, a note to the customer and an optional PDF — the fields exist because a buyer comparing five quotes is comparing five sets of assumptions, and the ones that are written down are the ones that survive.

11. Failure Modes

# · Failure · Mechanism · The arithmetic

1 · Margin added instead of divided · 20% markup = 16.7% margin · 3.3 points of turnover lost silently

2 · Rate inherited from an old rate card · Power, wages and capital all rose · ₹550 assumed vs ₹855 actual = 36% under-recovery

3 · Utilisation assumed at 100% · Denominator too large · 6,000 h assumed vs 3,600 actual = 17% low

4 · Material priced on finished mass · BTF ignored · ₹515 quoted vs ₹1,266 actual on this part

5 · Speed pushed for cycle time · Taylor's exponent · +25% speed = 59% of tool life at n = 0.25

6 · Small lot priced at large-lot rate · Setup not amortised correctly · ₹2,264 quoted vs ₹3,361 cost at 10-off

7 · Scrap priced at material cost · Value-added not counted · ₹1,266 assumed vs ₹1,844 actual per scrap

8 · Terms and escalation omitted · Working capital unpriced · 90 days at 13% = 3.21% of invoice

FAQ

What is a should-cost model, in one sentence?

A reconstruction of what a part should cost a competent supplier — material net of scrap credit, processing at a derived machine-hour rate, amortised setup, tooling, outside processing, yield loss, overhead and a stated margin — built from the drawing and the physics rather than read off a supplier's quote.

What is a fair machine hour rate for a CNC VMC in India?

A fully burdened rate of ₹500–900/hr for a three-axis VMC on general work is the market band, ₹1,000–1,500/hr for a large industrial VMC, and ₹1,200–2,500/hr for five-axis or high-precision work. The band is wide because utilisation is the dominant variable: the same ₹45 lakh machine costs ₹1,036/hr on a single shift and ₹855/hr on two, before any difference in wages or power. Always establish whether the quoted rate includes labour and margin.

Why does the same part cost 2.8× more for one piece than for six hundred?

Because setup does not scale. ₹6,100 of programming, fixturing and first-article inspection is 31% of the price at one piece and 0.4% at six hundred. The material and machining lines are nearly identical across the range; the entire difference is the amortisation of a fixed cost.

How do I estimate machining cost without a CAM system?

Use removal rate. Volume to remove divided by a conservative removal rate for the material and tool — 40–60 cm³/min in aluminium, 8–15 cm³/min in mild steel, 2–5 cm³/min in titanium — plus finishing time for the toleranced surfaces, plus 10–15% for tool changes and rapids. It will land within 20% of a CAM cycle time on prismatic parts, which is more than accurate enough for a quote that carries a 12% margin.

Does a tighter tolerance really cost more if the machine can hold it?

The machine can usually hold it; proving it is what costs. Below about ±0.05 mm the quote acquires a finishing pass, a temperature-stable environment, a CMM check and often a stress relief between roughing and finishing. That is a process change, not a parameter change, and it typically adds 15–40% to the machining line.

Should I quote the price including GST?

Quote the taxable value and state GST separately, at the rate applicable to the HSN code. Mixing tax into a unit price makes the quote non-comparable with every other quote the buyer receives and creates a reconciliation problem at invoice time. State the currency, the incoterm or delivery basis, the payment terms and the validity period on the same page.

How much does a bad quote cost?

On a ₹1,45,140 job, a 7.4% cost overrun from one material price error and a 5% rework rate is ₹10,740 — more than half the profit on the job, and more than the ₹6,750 of estimating cost that produced it. The failure is not usually a single mistake; it is two ordinary assumptions that were never checked.

What is buy-to-fly and why does it matter more than material price?

Buy-to-fly is purchased stock mass divided by finished part mass. It converts a per-kg price into an effective per-part cost, and it is the only way to compare a ₹560/kg aluminium plate against a ₹4,500/kg titanium bar fairly. On this bracket the ratio is 2.75:1; on an aerospace structural rib it can exceed 20:1, at which point the material line dominates the quote and near-net-shape processes become cheaper despite their tooling.

Related guides in this series: CNC Machining · CNC Feeds & Speeds · CNC Machining DFM · Injection Molding Economics · Print Farm Economics · Metrology & Dimensional Inspection · Tolerance Stackups · Lean Manufacturing.

References

[1] F. W. Taylor, On the Art of Cutting Metals, Transactions of the ASME, Vol. 28, 1907 — the original VT^n = C relation and the tool-life experiments behind the exponents used in §3. [2] ASM Handbook, Vol. 16: Machining, ASM International, 1989 — specific cutting energies, removal-rate practice and tool-life data by workpiece material. [3] Sandvik Coromant, Metal Cutting Technology: Technical Guide, current edition — recommended cutting speeds and feeds for 6061-T6 and Ti-6Al-4V, and the specific cutting force k_c model behind removal-rate estimates. [4] Y. Altintas, Manufacturing Automation: Metal Cutting Mechanics, Machine Tool Vibrations, and CNC Design, 2nd ed., Cambridge University Press, 2012 — cutting force coefficients, spindle power and the mechanics behind the removal-rate figures in §3. [5] ISO 8688-1:1989 and ISO 8688-2:1989, Tool life testing in milling — the standard conditions under which a tool-life exponent is measured, and the reason catalogue exponents differ between suppliers. [6] G. Boothroyd, P. Dewhurst and W. A. Knight, Product Design for Manufacture and Assembly, 3rd ed., CRC Press, 2010 — the should-cost structure of §1 and the machine-rate absorption method. [7] C. T. Horngren, S. M. Datar and M. V. Rajan, Cost Accounting: A Managerial Emphasis, 16th ed., Pearson, 2018 — machine-hour rate absorption, overhead allocation and the markup-versus-margin distinction. [8] AIAG & VDA, Measurement Systems Analysis (MSA) Reference Manual, 4th ed., 2010 — gauge R&R and the capability conventions; ISO 22514-2:2017, Statistical methods in process management — Capability and performance — Part 2: Process capability and performance of time-dependent process models — the C_{pk} definitions used in §6. [9] ISO 14253-1:2017, Geometrical product specifications (GPS) — Inspection by measurement of workpieces and measuring equipment — Part 1: Decision rules for verifying conformity or nonconformity with specifications — the measurement-uncertainty decision rules that make a tight tolerance expensive to prove. [10] The Micro, Small and Medium Enterprises Development Act, 2006 (Act 27 of 2006), Sections 15, 16, 19 and 23 — the 45-day payment cap, compound interest at three times the RBI bank rate with monthly rests, the 75% deposit for appeal, and the disallowance of such interest as a deduction. [11] MSME Samadhaan, Ministry of Micro, Small and Medium Enterprises, Government of India — the delayed-payment monitoring and Facilitation Council filing mechanism. [12] T. P. Wright, "Factors affecting the cost of airplanes," Journal of the Aeronautical Sciences, 3(4), 1936, pp. 122–128 — the learning-curve formulation C_n = C_1 n^{b} used in §4. [13] "Lower job order rates threaten small units' survival in Coimbatore," The Times of India — the reported gap between the ₹550/hr standard VMC rate and ₹250/hr being offered to job workers, cited in §2. [14] Indian job-work price observations, 2026: Type II sulphuric anodising at ₹150–250/kg on tonnage contracts and ₹0.25–1.00/in² for job work, with lot minimum charges of ₹1,200–2,500; CNC machining at ₹400–1,200/hr for three-axis work and ₹1,000–2,500/hr for five-axis, by cluster (Pune, Bengaluru, Hyderabad at the top of the band; Coimbatore, Chennai, Ahmedabad below) — the ranges in §2 and §7. [15] ISO 9001:2015, clause 8.5.1 and AS9102 First Article Inspection Requirement — the first-article inspection obligation that creates the ₹1,800 FAI line in the worked example.

Note on sources: the governing relations are the should-cost identity of §1, Taylor's VT^n = C for tool life, \dot V = a_e a_p v_f for removal rate, BTF = m_s/m_p for material, C_{good} = C_{attempt}/Y for yield, and C_{pk} capability for scrap; the commercial rules are the MSMED Act 2006 on payment terms and the markup-versus-margin identity. Every ₹ figure in this guide is derived from the parameter set stated in the text — a ₹45,00,000 VMC, 3,600 billable hours, ₹8.50/kWh, a ₹28,000 loaded operator wage, ₹560/kg 6061-T6 plate and ₹95/kg chip credit — and will move with those inputs. The method transfers; the numbers must be re-derived for a specific shop, city and year. A rate card is a hypothesis about costs. A should-cost model is a measurement of them.

More FabFlow blog posts